Solution of Large Scale Matrix Inversion on Cluster and Grid

Large scale matrix inversion has been used in many domains and block-based Gauss-Jordan (G-J) algorithm as a classical method of large matrix inversion has become the focus of many researchers. Many people show us their parallel version of G-J. But the large parallel granularity in those algorithms restricts the performance of parallel block-based G-J algorithm, especially in the cluster environment consisting of PCs or workstations. This paper presents a fine-grained parallel G-J algorithm to settle the problem presented above. Experiments are made based on YML a framework which enables using different middleware to make large scale parallel computing for its feathers of components reuse, easy programmability for non-computer professionals. Cluster and Grid environments are based on Grid'5000 platform, France. Experiments show us that the better performance of fine-grained parallel G-J algorithm and YML though overhead existing is a good solution for large scale parallel computing.